To Be Proactive or Not: A Framework to

Model Cyber Maneuvers for Critical Path

Protection in MANETs

Zhuo Lu

University of

Memphis

Lisa Marvel

Army Research

Laboratory

Cliff Wang

North Carolina State

University / Army

Research Office

Outline

• Background

– Cyber maneuvers in tactical MANETs

• Framework

– Models

– Optimization approach

• Evaluation and simulation results

• Conclusions

Tactical MANETs

• Mobile Ad-Hoc Network (MANET)

– infrastructure-less network of mobile wireless

devices for military operations

source

destination

Goals vs Issues in Cyber Missions• Issues / Constraints:

– Limited energy budget

– Limited power/bandwidth

– Distributed deployed in battlefields, may be easy to be

compromised by cyber attacks

– …

• To achieve successful army operations:

– Protecting a critical path

– Prolonging the network lifetime

– Securing critical nodes

– …

We must optimally design/coordinate cyber maneuvers to achieve security goals under constraints!

Cyber Maneuvers

• Cyber maneuver

– an action in the cyber space towards achieving the

goal in a mission

– e.g., software upgrade, patching, node

isolation/blocking, …

• Reactive or Proactive

– reactive: face security issues then solve!

• E.g., traditional intrusion detection

– proactive: prevent security issues from happening

(now and in the future)

• e.g., MTD.

Our Scenario and Objective• In a MANET deployed in adversary environment

– Nodes can be affected by virus from an attacker because of

new software vulnerability

– Goal: Make sure a critical path is always protected!

Should we be proactive or not?

• Be proactive: immediately patch a vulnerable node.

• Be reactive: patch a vulnerable node when it faces threats

… …

… …

… …

Source

Destination

infected nodes

infected node

Our Analytical Framework

• A new framework to model the effectiveness

and costs of cyber maneuvers, it integrates

– Network model

– Attack model

– Cyber maneuver model

– Cost model

– Optimization framework

Network and Attack Scenario• MANET: n nodes

– Node 1: the attacker that infect other nodes.

– Node n: patching node that performs a maneuver on nodes,

e.g., patching an infected node

• Assumption: patching node knows all info (e.g., node/link states)

– Nodes 2-n: legitimate nodes that can be infected, patched.

The patching node

(node n)

The attacker

(node 1)

… …

… … infection

process

maneuver

different link throughput

… …

Node States and Capabilities

Immune /

Patched

Quarantined

BlockedVulnerable

Susceptible

Infected

The capability of a node can be defined based on its state. E.g., capability= 0 if infected

Example: Node Capability

Positive values

Cyber Maneuvers

• Set of cyber maneuvers

– No Action

– Patch Completely upgrade a node’s software

– Software Heal partly recover the routing

function

– Node Block completely disable a node

The patching node

(node n)

maneuver

Node State Transition

Immune /

Patched

Quarantined

Blocked

Vulnerable

Susceptible

Infected

new software exploit

infected node nearby

infected

Node block

Software heal

Patch

reactive

proactive

Q: Should we be proactive or reactive?

Cost Model

• Energy Cost in MANETs

– A cyber maneuver costs energy at all involved

nodes.

• Patch > Software Heal > Node Block > No Action

The patching node

(node n)

node i

wants to maneuver node i

node j2

node j3

node jK-1

jK=i

j1=n

K nodes on the maneuver path

Optimization Goals

• Lots of objectives, e.g.:

– All nodes on the path must not be infected;

– The overall capability of the path (i.e., the sum of

capabilities of all nodes on the path) should be

maximized;

– The overall capability of the network (i.e., the

sum of capabilities of all nodes in the network)

should be maximized;

– The cost to protect such a path should be

minimized.

• Cannot be all met at the same time!

Our Strategy

• Maximize one objective: (primary focus)

– maximize the lifetime of a critical path

• Add multiple constraints:

– E.g., the total capability in the network, on the

path, the cost of the maneuver.

• Based on two views:

– Current view (cannot predict the future).

– Statistical view (can somehow predict the future)

• e.g., statistical consumption of energy, node mobility, …

Our Formulation I• Based on current view of the network: have all node info (e.g., remaining

energy, link rate), but no future info (e.g., who nodes will move)

node x1

(source) node xY

(destination)

critical path

node x2

node x3

node xY-1

Maximize the minimum

energy on the path (so

maximizing the lifetime

of the critical path

All nodes on the critical

path must be in good

statesThe total capability on

the path must be large

enoughThe total capability in

the network must be

large enoughAll nodes involved in a

cyber maneuver must

have enough energy.

Solution• Indications:

– defer cyber maneuver

(i.e., choose No Action)

as much as possible

unless we have to act

(when the constraints

do not hold)

– Because we only have

the current view, and

cannot predict the

future.

– Try not to be proactive

unless we have to!

Formulation II• Based on statistical view of the network:

– Node distribution, mobility statistics, energy consumption

node x1

(source) node xY

(destination)

critical path

node x2

node x3

node xY-1

Maximize the probability that

there still exists a secure critical

path after time duration 𝜏

Solution: Sufficient information gives us the best proactive solutions!

Simulation Setups• A MANET:

– Network size: a 1000-meter by 1000-meter region.

– Node setups: Transmission range of 100 meters, uniformly

distributed with independent mobility.

– Energy mode: the energy consumption is a linear function

of the number of traffic transmissions of each node.

– Critical path: we randomly choose two nodes as the source

and the destination

– Attack and defense:

• There exists an adversary in the network that attempts to infect

other nodes as long as they meet.

• The patching node aiming to make the best decision to maneuver

other nodes in the network in order to maximize the lifetime of the

critical path between the source and the destination.

Node States and Maneuvers

Result I

• Average capability on the critical path

optimization based on current view

optimization based on statistical info

Result II

Result III

Result IV

• If there is an error in statistic info

(10% error)

Shoter than lifetime optimized based on

current view

Conclusions• A framework to model cyber maneuvers

– Easily adopt more node states, maneuvers, cost models, …

• Accurate statistical info is a key enabler for proactive

cyber maneuvers for critical path protection

– If we only have current view, defer proactive strategies

– If we have sufficient statistical info, choose the best

proactive strategies based on the optimization framework.

– Wrong statistical info may lead to worse performance

• More to improve:

– Information collection.

– Trust on a path.

– Fine-grained statistical error analysis.